Coverage quality
A transparent audit of source coverage, PDF status, and whether papers have usable full text, summaries, abstracts, or no usable text.
1699
papers tracked
77.9%
complete full-text coverage
2
suspect full-text records
Text basis
This is the evidence basis available to summarization and synthesis: complete extracted full text when available, otherwise summary or abstract fallbacks.
| Text basis | Papers | Share |
|---|---|---|
| Complete full text | 1323 | 77.9% |
| Suspect or incomplete full text | 2 | 0.1% |
| Summary only | 374 | 22.0% |
| Abstract only | 0 | 0.0% |
| No usable text | 0 | 0.0% |
PDF status
| Status | Papers | Share |
|---|---|---|
| Downloaded | 1325 | 78.0% |
| Paywall | 185 | 10.9% |
| Not Found | 90 | 5.3% |
| Pending | 81 | 4.8% |
| Error | 18 | 1.1% |
Source coverage
| Source | Papers | Full text | Assessed |
|---|---|---|---|
| arxiv | 742 | 99.7% | 100.0% |
| openalex | 621 | 69.9% | 100.0% |
| semantic_scholar | 327 | 42.8% | 100.0% |
| manual | 9 | 100.0% | 100.0% |
What counts as full text?
A paper is counted as complete full text only when extracted text is long enough and does not contain known truncation or extraction-error markers. Short, noisy, or marker-laden text is flagged as suspect so it is not treated as reliable full-text coverage.